Citation Context

...within vocal parts or parts that differ in instrumentation but are otherwise harmonically similar. Then we compute the amplitude spectra of the lengthequalized measures using a Fast Fourier Transform =-=[CT65]-=-. The logarithmic sound pressure level measured in decibels is given by p Lp[dB] = 20 · log10 . p0 The A-Weighting curve is now applied to the sound pressure level, L[dB(A)] = Lp[dB] + A( f ) , and th...

Citation Context

...otic systems and was later introduced in the domain of music analysis for visualization of an audio recording’s time structure [Foo99]. We compute the selfsimilarity via the Bray-Curtis dissimilarity =-=[BC57]-=-, d(u,v) = ∑i |ui − vi| ∑i |ui + vi| , (1) as distance function. It is one of the most well-known nonmetric ways of quantifying the difference between data sets and delivers robust and reliable dissim...

Citation Context

...h larger due to their employed song segmentation strategy. Our measure-wise segmentation allows us to employ a belief propagation algorithm instead. Belief propagation [YFW03,TF03], first proposed by =-=[Pea82]-=-, is a message passing algorithm performing inference on graphical models like Markov random fields (MRF) or Bayesian networks which have been successfully employed in artificial intelligence and info...

Citation Context

...3,...,n. This concept was already employed by [EKR87] for analysis of chaotic systems and was later introduced in the domain of music analysis for visualization of an audio recording’s time structure =-=[Foo99]-=-. We compute the selfsimilarity via the Bray-Curtis dissimilarity [BC57], d(u,v) = ∑i |ui − vi| ∑i |ui + vi| , (1) as distance function. It is one of the most well-known nonmetric ways of quantifying ...

Citation Context

...xn), compare its elements pairwise according to some distance function d, and store the results in a self-similarity matrix Ci, j = d(xi,xj) : i, j ∈ 1,2,3,...,n. This concept was already employed by =-=[EKR87]-=- for analysis of chaotic systems and was later introduced in the domain of music analysis for visualization of an audio recording’s time structure [Foo99]. We compute the selfsimilarity via the Bray-C...

Citation Context

...nce between data sets and delivers robust and reliable dissimilarity results throughout many applications. As the analysis window for computing measure-boundary similarity, we employ a Hanning window =-=[BT59]-=- centered at measure boundaries, Figure 2. The Hanning window, defined as w(n) = 1 ( ( )) 2πn 1 − cos , (2) 2 N − 1 smoothly transitions from one in the center to zero at the borders. This ensures tha...

Citation Context

... curves [Lin]: A-weighting (blue), B (yellow), C (red), D (black). for normal environments up to D for loud aircraft noise, Figure 3. It is derived from the 40-phon equal loudness contour proposed by =-=[FM33]-=- as an approximation of its inversion to resemble gain. Figure 4: Self-similarity matrix Cli−1,li of an isolated song part. Blue marks regions of high similarity, while regions of low similarity are i...

Citation Context

...ld lose synchronization with the song over time, leading to noticeable artifacts at the measure boundaries during resynthesis. After evaluating the performance of various beat trackers, like BeatRoot =-=[Dix07]-=-, beatsync [Che02], BTrack [SDP09], B-Keeper [RP07], and the tracking components available for the Sonic Visualizer [CLSB06] software, we selected the [aufTAKT] tempo and beat tracking system by zplan...

Citation Context

...is. After evaluating the performance of various beat trackers, like BeatRoot [Dix07], beatsync [Che02], BTrack [SDP09], B-Keeper [RP07], and the tracking components available for the Sonic Visualizer =-=[CLSB06]-=- software, we selected the [aufTAKT] tempo and beat tracking system by zplane.development [Zpl13]. It was the only beat tracker able to reliably track the beat signal in our experiments and has been f...

Citation Context

...basic tools necessary for machine aided understanding and restructing of musical pieces. In our work, we focus on music that can be analyzed and segmented with current technologies like beat trackers =-=[Ros92]-=- and structure analysis tools [Sch06]. The analysis and synthesis of sound consisting of textured patterns without defined pitch, rhythm, dynamics and timbral qualities (e.g., [LWZ04]) is beyond the s...

Citation Context

...ed understanding and restructing of musical pieces. In our work, we focus on music that can be analyzed and segmented with current technologies like beat trackers [Ros92] and structure analysis tools =-=[Sch06]-=-. The analysis and synthesis of sound consisting of textured patterns without defined pitch, rhythm, dynamics and timbral qualities (e.g., [LWZ04]) is beyond the scope of this paper. 2. Related Work T...

Citation Context

...ding to noticeable artifacts at the measure boundaries during resynthesis. After evaluating the performance of various beat trackers, like BeatRoot [Dix07], beatsync [Che02], BTrack [SDP09], B-Keeper =-=[RP07]-=-, and the tracking components available for the Sonic Visualizer [CLSB06] software, we selected the [aufTAKT] tempo and beat tracking system by zplane.development [Zpl13]. It was the only beat tracker...

Citation Context

...ong over time, leading to noticeable artifacts at the measure boundaries during resynthesis. After evaluating the performance of various beat trackers, like BeatRoot [Dix07], beatsync [Che02], BTrack =-=[SDP09]-=-, B-Keeper [RP07], and the tracking components available for the Sonic Visualizer [CLSB06] software, we selected the [aufTAKT] tempo and beat tracking system by zplane.development [Zpl13]. It was the ...

Citation Context

...ke beat trackers [Ros92] and structure analysis tools [Sch06]. The analysis and synthesis of sound consisting of textured patterns without defined pitch, rhythm, dynamics and timbral qualities (e.g., =-=[LWZ04]-=-) is beyond the scope of this paper. 2. Related Work The contemporary concatenative audio resynthesis approach of cutting and stitching started in the 1950s and has been the c○ The Eurographics Associ...

Citation Context

...R03]. Audio textures [LLW ∗ 02] segment input audio into short clips that are subsequently rearranged. Similar to concatenative audio resynthesis, these methods are typically not applicable to music. =-=[SERlG06]-=- summarize similar sound texture generation methods. Concatenative audio synthesis [Sch06, S ∗ 00] uses databases of sound snippets to assemble a specific target sound. While this approach has been us...

Citation Context

...ith only a few jumps left that a trained c○ The Eurographics Association 2013.J.-P. Tauscher et al. / Audio Resynthesis on the Dancefloor 70 In the future, automatic recognition of musical structure =-=[PMK10]-=- could be employed to simplify the annotation process and to provide a more intuitive interface to our algorithm. Music structure analysis is still an active area of reserch, but even today existing m...

Citation Context

... pieces. Audio and user directed sound synthesis calculates the self-similarity of an audio source per frame and has been successfully tested on stochastic or periodic sound patterns but not on music =-=[CBR03]-=-. Audio textures [LLW ∗ 02] segment input audio into short clips that are subsequently rearranged. Similar to concatenative audio resynthesis, these methods are typically not applicable to music. [SER...

Citation Context

...cting sound textures and sound scapes and does not perform well on musical sources. In order to make audio resynthesis techniques applicable to music and other audio containing large-scale structure, =-=[WM11]-=- introduced a multi-resolution scheme for fast selfsimilarity analysis capturing the sonic source from multiple measures down to single samples, allowing perfect alignment of the cuts without blending...

Citation Context

...the input song, Figure 4. 12200 A( f ) = 2 · f 4 ( f 2 + 20.62 )( f 2 + 122002 ) 1 · √ ( f 2 + 107.72 )( f 2 + 737.92 ) Loudness perception generally is a much more complex task than just A-Weighting =-=[Ols72]-=-, but it delivers a good approximation sufficient for our application. We therefore apply the A-Weighting to our analysis window with the same setup as in the previous section. This accounts for good ...

Citation Context

...heme for fast selfsimilarity analysis capturing the sonic source from multiple measures down to single samples, allowing perfect alignment of the cuts without blending or scaling. A genetic algorithm =-=[WM12]-=- was subsequently proposed to rearrange the segments into soundtracks subject to user-specified constraints. This approach performs reasonably well even on structured music, but often exhibits noticea...

Citation Context

...ation with the song over time, leading to noticeable artifacts at the measure boundaries during resynthesis. After evaluating the performance of various beat trackers, like BeatRoot [Dix07], beatsync =-=[Che02]-=-, BTrack [SDP09], B-Keeper [RP07], and the tracking components available for the Sonic Visualizer [CLSB06] software, we selected the [aufTAKT] tempo and beat tracking system by zplane.development [Zpl...